The InfoQ Podcast

AI Autonomy Is Redefining Architecture: Boundaries Now Matter Most

13 snips
Mar 4, 2026
Jesper Logren, an enterprise architect leading generative AI work at DXC and author of Design or Be Designed, discusses why autonomy changes architecture. He explains how autonomous systems drift, why retrofitting AI fails, and why defining strict boundaries, joined design and governance, and clear agent guardrails are now essential.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
INSIGHT

A Maturity Ladder For Agentic Systems

  • AI maturity spans from ad hoc assistants to multi-agent autonomy.
  • Jesper defines levels 1–5 and inserts 2.5 (multi-agent without autonomy) to show big deltas when moving to true autonomy.
ADVICE

Specify Agent Decision Rights

  • Explicitly set authority and decision rights for every agent.
  • Jesper stresses clarity on which decisions an agent may make to avoid unexpected, high-risk outcomes as autonomy increases.
ADVICE

Use Seven Boundary Dimensions To Contain Agents

  • Use seven boundary dimensions (goals, scope, risk, semantics, evidence, authority, interfaces) to contain agents.
  • Jesper gives scope as an example, insisting on explicit interface rules for ERPs/CRMs to avoid leaks.
Get the Snipd Podcast app to discover more snips from this episode
Get the app